论文部分内容阅读
§11-3梯度法梯度法的优点是收敛性好,有利于电路结构的最优化。其缺点是需要计算梯变(参见第七章),程序编制较复杂。1.最速下降法梯度法中最简单最易理解的是最速下降法。此法的特点是按照误差函数的负梯度方向修正设计参数向量X 值,使之在误差空间中逐步逼近极小值。某一点的负梯度方向是该点误差函数值下降最快的方向。
§11-3 gradient gradient method has the advantages of good convergence, is conducive to the optimization of the circuit structure. The disadvantage is the need to calculate the ladder (see Chapter VII), programming is more complicated. 1. steepest descent gradient method is the easiest and most understandable is the steepest descent method. The characteristic of this method is that the X value of the design parameter vector is corrected according to the negative gradient of the error function so as to gradually approach the minimum value in the error space. The negative gradient direction at a certain point is the direction of the fastest drop of the error function value at this point.